COMSOL Multiphysics^® 6.2 Release Highlights

Uncertainty Quantification Updates

For users of the Uncertainty Quantification Module, COMSOL Multiphysics^® version 6.2 introduces specification of correlated input parameters via a correlation matrix, as well as autogeneration of trained surrogate functions and new models for inverse uncertainty quantification and the manufacturing of solidly mounted resonators. Learn about these updates and more below.

Dependent Input Parameters

You can now define correlated input parameters by grouping them into correlation groups. The data points are sampled with the Gaussian copula method, which uses the correlation matrix and marginal distributions for data sampling. When surrogate function settings are used to specify the input parameters and verification input parameters of the Monte Carlo analysis, the same correlation settings used in training the surrogate function are then used to sample data for the Monte Carlo analysis and verification. These operations expand the functionality of uncertainty quantification studies to better support multivariate input parameters.

Add Surrogate Function

The surrogate-based uncertainty quantification study types — including Sensitivity Analysis; Uncertainty Propagation; Reliability analysis, EGRA; and Inverse uncertainty quantification — now create global surrogate functions after the study is computed. The surrogate functions can then be evaluated with or without the Uncertainty Quantification Module.

Error Region Plots

In addition to using the previously available Error Bar plot, you can now also visualize uncertainties as a colored region within the plot. This new addition can be used to visualize, for example, the uncertainty in the Gaussian process regression. The displayed uncertainty level can also be customized to visualize, for example, the region for 1 sigma, 2 sigma, and so forth.

New Tutorial Models

COMSOL Multiphysics^® version 6.2 brings two new tutorial models to the Uncertainty Quantification Module.